The Google Analytics
migration atlas.

Here is the number that surprises most teams I talk to. In my effort estimator, a typical setup (30 custom events, 5 custom dimensions, one property, two people) needs about 81 hours of hands-on work to move from GA4 to Plausible.

That is about one working week for two people. The calendar still reads two to three weeks, because the parallel run takes 14 days however fast the team works.

I sort every route on this site into three families. Inside a family the work looks alike, so this table is the short version of the atlas.

Cookieless dashboards fit content sites, docs and marketing pages, where the questions are pages, sources and a handful of goals. Most of the work is mapping goals. The 14 days are there so you see two full weekly cycles before you trust the new numbers.

Full analytics suites are for teams that open GA4's deeper reports every week: ecommerce, segments, custom dimensions. Matomo can also run on your own servers. The extra time goes into the history import on Matomo and into the Tag Manager rebuild on Piwik PRO.

Product analytics answers a different question. Reports start from a signed-in person (activation, retention, feature use), so you rebuild identity along with events. If your reports are about traffic, this family adds weeks you do not need.

Removing the tag deletes nothing. Google's retention setting does not touch standard aggregated reports, so the old property stays readable for as long as you keep it. Explorations and funnel reports are the exception. On a standard property they reach back only 2 or 14 months, depending on the setting.

Three destinations can pull GA4 history into their own dashboards. All three take aggregated numbers, never individual visits.

• Plausible imports daily totals from your first GA4 visitor up to your first Plausible visitor. Revenue does not come over, and goal numbers appear only after you recreate the goals.
• Fathom imports daily data but no events. A large import can take up to 48 hours.
• Matomo imports aggregated reports without the visitor log. Google's daily API limits can stretch a big import over several days.

The shared loss is the ability to slice old data. I treat imported history as a read-only chart, because it cannot be segmented or filtered the way new data can. On Plausible and Matomo, unique visitors over weeks and months also drift from GA4, since only daily figures travel.

The other routes start with an empty chart. Mixpanel and Amplitude can backfill events, but only from GA4's BigQuery export. Mixpanel suggests a year of history or less and warns that backfill can affect billing. My rule for every route: decide how many months you need before you delete the GA4 property, not after.

I would stay, or keep GA4 on part of the site, in four cases.

• Google Ads runs on GA4 data. A linked property sends key events to Google Ads as conversions and shares audiences for remarketing. Cut the link and those conversions and audiences stop updating.
• You need credit for assist channels. Multi-touch attribution for paid campaigns stays in GA4 or moves to Matomo. The cookieless dashboards do not offer it.
• The cookie banner is your only reason. If Google Ads or Meta tags stay on the site, they still need consent in the EU, so the banner stays too.
• Nobody owns the parallel run. Fourteen days of daily comparison needs one named person. Without one, postpone the move instead of switching blind.

A split setup is normal: GA4 on the pages that feed Ads, the new tool everywhere else. If you still cannot pick a family, the five questions on the GA4 exit page take ten minutes.